International Journal of Advances in Applied Sciences (IJAAS)
Vol. 3, No. 1, March 2014, pp. 25~32 ISSN: 2252-881425 Journal
homepage: http://iaesjournal.com/online/index.php/IJAAS Modeling
the Multiple Indirect Effects among Latent Constructs By Using
Structural Equation Modeling: Volunteerism Program Wan Mohamad
Asyraf Bin Wan Afthanorhan Department of Mathematics, Faculty of
Science and Technology, Universiti Malaysia Terengganu, 21030 Kuala
Terengganu, Malaysia Article InfoABSTRACT Article history: Received
Jul 27, 2013 Revised Feb 23, 2014 Accepted Mar 3, 2014 This study
aimed to evaluate the factors used for develop a best model of
multiple indirect effect among latent constructs by using
Structural Equation Modelling (SEM) on volunteerism program as a
research subject. The data is
collectedthroughquestionnairesdistributedatfourhighereducation
institution. This questionnaire is constructed based on four
dimensions which
aremotivation,benefits,govermentsupport,andbarrier.Thedatawere
distributedbyusingstratifiedsamplingtechniqueandinvolving453
respodents. In this case, the data were analyzed by using Analysis
Moment of Structural (AMOS) 18.0 in order to examine the influence
of exogenous and
endogenousvariables.Asaresultshowedthatthegovermentsupportis
significantanddirectinfluencesonmotivation,benefits,andbarrier.
Moreover,thebenefitsandbarrierissignificantanddirectinfluenceon
motivation. In generals, the findings revealed that benefits
influence is most crucial for motivation of volunteerism.Keyword:
Analysis of Moment Structural (AMOS) Stratified Sampling Technique
Volunteerism Structural Equation Modelling Multiple IndirectEffect
Copyright 2014 Institute of Advanced Engineering and Science. All
rights reserved. Corresponding Author: Wan Mohamad AsyrafBin Wan
Afthanorhan, Department of Mathematics, Faculty of Science and
Technology,Universiti Malaysia Terengganu,21030 Kuala Terengganu,
Malaysia.Email: [email protected] 1.INTRODUCTIONThis study
emphasizes the level of involvement in volunteerism program
especially among youth from higher education institution chosen.
One of these factors can be examined by the reason of volunteer
which is
consideredasmotivation(Rhyne,1995).Theotherthreevariablesincludearebenefits,barrier,and
govermentsupport.Allofthesefactorsisregarding ontheliteraturereview
previously.Volunteerismis defined as a professional or
non-professional person who provides a service to a welfare or
development organization, usually without reimbursement (The White
Paper for Social Welfare, 1998). Barrier is referred as not about
supported volunteering specifically (Eva Schindler-Raiman, 1987).
According to (Dingle, 2001), the benefits is extremely important if
had supported by the contribution of goverment. Thus, this barrier
hinders the growth of volunteery activities. In this study, the
benefits, and barrier play a role as mediator variable since these
variables can become exogenous and endogenous variable
simultaneously.Therefore, the prior studies is to examine the
relationship and influence between goverment support, benefits, and
barrier on motivation as well as their different relationships. In
generals, this study employs the indirect effect in order to
achieve the objective research and research question. The indirect
effect can be namely as the mediating effect or intervening effect.
As usual, the Confirmatory Factor Analysis (CFA) procedure should
be applied in order to achieve the reliability and
validityofmeasurementmodel.Accordingto(Hairet.al,2006)explainCFAcanbenamelyasthe
ISSN: 2252-8814 IJ AASVol. 3, No. 1, March 2014 :25 32
26measurementmodel.InStructuralEquationModelling(SEM),therearetwotypesofmodelwhichis
measurement model and structural model.The measurement model is
frequently used nowadays among researchers to undergo the CFA
procedure. In this case, this study applies CFA procedure before
furthering the structural model in order to achieve the validity of
latent contructs. In addition, measurement model can be known as
model hypothesis testing in
ordertoobtaintheestimationmodelwithmorefit.Firstandforemost,theunidimensionalityprocedure
shouldbeapplyforwholemeasurementmodeltoremovethemeasuringitemsthathavethelower
standardized factor loadings (0.90,NFI>0.90,AGFI>0.90andan
insignificant c2, to show unidimensionality. In addition,
itemloadings shouldbe above 0.70, to show that over half the
variance is capturedbythelatentconstruct(Chin,
1998,Hairet.al,1998,Segars,1997, Thompson et.al, 1995) Discriminant
ValidityCFA used in Covariance-Based SEM
onlyComparingthec2oftheoriginalmodel withanalternativemodelwherethe
construct. If the c2is significantly smaller in the original model,
discriminant validity has been shown (segars, 1997) Convergent and
Discriminant ValidityPCA used in PLS can assess factor analysis but
not as rigorously as a CFA in LISREL does and without examining
unidimensionality Each construct AVE should be larger than
itscorrelationwithotherconstructsand each items should load more
highly on its assignedconstructthanontheother constructs
Reliability Internal ConsistencyCronbach AlphaCronbach alpha should
be above 0.60 for explanatoryresearchandabove0.70for
confirmatoryresearch(Nunally,1967, Nunally, 1978, Nunally &
Bernstein, 1994, Peter, 1979) SEMThe internal consistency
coefficient should be above 0.70 (Hair et.al., 1998, Thompson
et.al, 1995) Unidimensionality ReliabilityCovariance-Based SEM
onlyModel comparison favor unidimensionality
withasignificantlysmallerc2inthe proposedmeasurementmodelin
comparisonwithalternativemeasurement model (Segars, 1997) Then, the
model modification verification should be attempted .This is
because this procedure can remedy the multicollnearity problem.
Based on the statistics assumption, the error should be
uncorrelated or indepedently. According to (Alias Lazim, 2011)
explain when more than one indepedent variable apperas in
modelling, it is possibe that these variables are related to eash
other. Means that, the multicollinearity among variables or
constructs is said to be exist.Thus, the constraints or double
headed arrow should be employed.
TheresearcherscancovarytheerrorbasedonthemodificationindicespresentinAnalysisMomentof
Structural (AMOS) output. The acceptance model when the contraints
applied on the same factor. According to Zainudin, 2012 explain the
error should be correlated when the covariance present value
greater than 15. However, Byrne, 2010 suggest the covariance should
be applied when the valus is greater than 10. Hence, it depends on
the researchers to apply the constraint based on ther literature
supported. The Table 2 presented shows the type of fitness indexes
with literature supported. 3.RESULTS AND ANALYSIS
StructuralEquationModelling(SEM)hastwotypesofmodelwhichismeasurementmodeland
structural model. Basically, mesurement model is frequently used
nowadys among reseracher to analyze for ISSN: 2252-8814 IJ AASVol.
3, No. 1, March 2014 :25 32 28Confirmatory Factor Analysis (CFA).
Hence, the researcher needs to run CFA procedures for each
construct
involvedinthestudy.Allmeasurementmodelsmustbevalidatedandacceptedpriortomodellingthe
structural model. In this case, there are have 4 dimension which is
motivation (16 items), goverment support
(9items),barrier(8items),andbenefits(14items).Accordingto(Hairet.al,2010)explainthefactor
loadings for each items should be greater than 0.6. However, factor
loading which greater than 0.50 is also accepted depend on the
decison by the researcher if have strong reason not to do so. The
Table 3 shows the territory items results leave after remove. Table
2. Type of fitness Number of Category Name of Index Index Full Name
Level of Acceptance Literature Absolute Fit GFIGoodness of Fit
IndexGFI>0.90J oreskog and Sorbom(1986) AGFIAdjusted Goodness of
Fit AGFI>0.90J oreskog and Sorbom(1986) SRMRStandardized Root
Mean Square Error Approximation SRMR0.95Tucker and Lewis (1973)
RNIRelative Noncentrality Index RNI>0.90McDonald & Marsh
(1990) CFIComparative Fit IndexCFI>0.95Bentler (1989,1990)
IFIIncremental Fit IndexIFI>0.90Bollen (1989) CommentHigher
values of incremental fit indices larger improvement over the
baseline model fit Parsiminous FitChisquare/ DFChisquare Degree of
FreedomChisq/ DF
